Chow, C.W., Urquhart, B., Lave, M., Bosch, J.L., Kleissl, J.

Solar forecasting mitigates issues with the variability of the solar resource. For intra-hour forecasting, a technique based on a binary cloud map from Total Sky Imager has been developed at UC San Diego. While most clouds cause rapid and large ramps in solar irradiance on ground, the cloud types strongly affect the magnitude of the ramp based on the cloud optical depth (Fig. 1). The ability to estimate cloud optical depth from sky imagery alone will improve solar forecasts.

In this study, sky image textural features are extracted using a gray level co-occurrence matrix and images are classified into groups according to the clear sky index kt computed from the Global Horizontal Irradiance (GHI) data. The training data set is generated based upon selected sky images sequences with known ranges of kt. A k-nearest neighborhood (kNN) classifier is chosen to classify test images based on textural features and validation is conducted through comparing the kt of test images to the kt of the classified group. Applications to determine other solar variability measures will also be discussed.

Figure 1. Global Horizontal Irradiance (GHI) data and corresponding cloud group. Cloud group in (a) shows less solar irradiance is attenuated than the cloud group in (b).